Research map / Label-efficient and robust learning / Domain adaptation and generalization
Render-to-Adapt: Unsupervised Personal Adaptation for Gaze Estimation
Yangshi Ge, Zheng Liu, Feng Lu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add iris task during pre-training (oma) (ablation) | ResNet-18 gaze estimator · DE→DM | angular error (deg): 8.75 -> 7.23 | Improved |
| add iris pre-finetuning on rendered images (js) (ablation) | ResNet-18 + oma · DE→DM | angular error (deg): 7.23 -> 6.29 | Improved |
| add self-supervised render-cycle calibration loop (sg) (ablation) | ResNet-18 + oma + js · DE→DM | angular error (deg): 6.29 -> 4.84 | Improved |
| add Lcycle on top of Lanchor (ablation) | R2A · DE→DM | angular error (deg): 6.29 -> 4.84 | Improved |
| add Lanchor on top of Lcycle (ablation) | R2A · DE→DM | angular error (deg): 5.21 -> 4.84 | Improved |
| increase unlabeled calibration images from 1 to 10 (ablation) | R2A · DE→DM | angular error (deg): 5.78 -> 4.84 | Improved |
| increase unlabeled calibration images from 10 to 20 (ablation) | R2A · DE→DM | angular error (deg): 4.84 -> 4.77 | No clear effect |
| use GazeGaussian instead of GazeNeRF as the differentiable renderer (ablation) | R2A · DE→DM | angular error (deg): 5.13 -> 4.84 | Improved |
| use ResNet-50 backbone instead of ResNet-18 (ablation) | R2A · DE→DM | angular error (deg): 4.84 -> 4.64 | Improved |
| apply UDA methods (PnP-GA, UnReGA) per individual instead of on the population | UnReGA · DE→DM | angular error (deg): 5.11 -> 5.72 | Hurt |
| R2A unsupervised versus supervised personal fine-tuning | R2A · DG→DM | angular error (deg): 5.38 -> 4.92 | Improved |
Closest papers by meaning
- Improving Few-Shot User-Specific Gaze Adaptation via Gaze Redirection Synthesis (CVPR 2019 · similarity 0.91)
- Alfa: Attentive Low-Rank Filter Adaptation for Structure-Aware Cross-Domain Personalized Gaze Estimation (AAAI 2026 · similarity 0.91)
- Gaze from Origin: Learning for Generalized Gaze Estimation by Embedding the Gaze Frontalization Process (AAAI 2024 · similarity 0.89)
- UVAGaze: Unsupervised 1-to-2 Views Adaptation for Gaze Estimation (AAAI 2024 · similarity 0.89)
- Enhancing 3D Gaze Estimation in the Wild using Weak Supervision with Gaze Following Labels (CVPR 2025 · similarity 0.88)
- OmniGaze: Reward-inspired Generalizable Gaze Estimation in the Wild (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
